Never has there been so great an appetite for data-driven decision making across our government customers. If done in the right way, data science can help make effective use of scarce resources, identify and prioritise threats, optimise operations and enable critical decisions to be reached in a timely manner. However, driving value from your data is not as simple as throwing it into a data visualisation tool and creating an array of aesthetically pleasing charts to present to your stakeholders; it requires careful planning, detailed understanding of the data you have (and do not have) and, most importantly, a clear set of hypotheses that you want to test with your data.
Similarly, there needs to be upfront consideration of what you will do with the outputs and insights that you gather from your data analysis. There is little benefit in having a dashboard that identifies areas for improvement without a business change strategy that will allow you to implement changes based on what the data is telling you. Who is going to review the outputs? How often will they do this? With whom do they share any insights? Who is going to turn these insights into tangible, operational improvements? Developing this thinking early and alongside the data analysis will be critical in the journey to harnessing the power of your data and truly making data-driven decisions.
With this in mind, Chris Sweeney shares his approach to ensuring that data analytics and business analysis are considered simultaneously to enable organisations to both identify and deliver tangible benefits from the data that they hold.
High level assessment
The first stage is an opportunity assessment; the purpose of this is to join a top-down business analysis exercise to identify the high level hypotheses you are hoping to test with a bottom-up data assessment to understand what data is available and where it is stored. This lets you form an initial opinion as to which hypotheses you are likely to be able to test with the data that is available.
Using this information, we then create a prioritised list of hypotheses to be tested in the next phase. The hypotheses are likely to be a mix of data-driven (i.e. what interesting patterns, insights and trends can we identify in the data that were previously unknown) or user-driven hypotheses (i.e. where do the stakeholders think there are likely to be opportunities but have not previously been able to prove their case without supporting evidence). Ensuring there is buy-in to act on the potential opportunity is also key and should be part of your initial checkpoint before deciding to progress into a more detailed assessment.
Detailed assessment
Having prioritised the list of hypotheses and therefore likely datasets we will need to use, the next stage involves an iterative process to test each hypothesis:
1. Identify and cleanse data – for each hypothesis we determine the datasets and fields that will be required to run the analysis needed to either prove or disprove the hypothesis. We then clean, standardise and enrich the fields required.
2. Analyse – we then run rapid analysis to test the key metrics that would prove or disprove the hypothesis. There are two possible outcomes at this stage:
The hypothesis is proven to be true and relevant insights are identified from the data warranting further investigation and development into visualisations or other outputs.
The hypothesis is proven to be false, and no relevant insights are found meaning we can quickly move onto the next hypothesis, factoring in any learnings.
3. Visualise – we then identify how the data can be most effectively visualised and presented (i.e. what types of charts, graphs, maps, clustering, network diagrams etc. most easily allow the insights to be visualised and interpreted). It is important to ensure that all analysis is explainable and presented at an appropriate level of detail for decision-makers.
4. Validate insights – we share the results of the analysis and visualisations with key stakeholders to validate the insights. This validation is critical to identify any data quality issues and check whether the insights identified based on the data under analysis are still likely to be relevant against any significant business or operational changes and drive forward the iterative analysis process.
We then develop these insights to drive a business change strategy. This involves prioritising the insights that have been uncovered during this iterative process, agreeing what changes the organisation will make in order to leverage the insights identified and agreeing how these changes can be implemented. We work closely with the stakeholders to ensure that we have considered all of the challenges and barriers in implementing any change, and continue to support the organisation through into exploitation.
Exploitation
Effective exploitation of the insights gained from the analytics process is likely to involve change. Without the appetite to change and a culture that is willing to support difficult decision making on the back of new insights, it’s hard to guarantee return on your analytics investment. While it may feel like the hard work has been done in having cleansed, merged and analysed your data, this is where to take the next step of creating actionable insights and it becomes possible to act on insight and see benefits realised. This is where our work during the initial opportunity assessment stage, where we gain commitment to act on insights found, is critical. Without that understanding and commitment upfront, we can be left with an interesting insight, but no viable course of action.
However, the exploitation phase is not just about taking one-off action based on what you have learnt. While you may realise benefit in the short term, without a longer term plan for both business change and a robust technical platform and repeatable data management and analysis processes, you risk diminishing returns on your initial investment. Planning to productionise your data analysis processes can avoid technical debt from tactical or one off activities and support a virtuous cycle of informed decision making.
In summary, this approach helps us support our customers across their complex data landscape where they may not know where to begin – or where their data analytics may be taking them. From Chris’s experience, understanding the business drivers, barriers to change and how data can drive decisions is just as important as getting the right machine learning tools and statistical techniques in place. Truly becoming a data-driven organisation is about far more than just the data.
Comments